AI Chatbots
I build assistants that answer visitors' questions from your own approved content, say plainly when they do not know, and hand the conversation to a person when they should.
New v6 draft from the service brief, awaiting review
The problem
- Visitors ask the same questions by email and phone, and the answers are already somewhere on your site.
- You have seen chatbots invent answers, and you do not want one speaking for your business that way.
- You are not sure what visitors would type into a chatbot, or where those conversations would be stored.
What I do about it
- I limit the assistant to the content you approve, and test it on real questions before it goes live.
- I set it to say when it does not know and to offer a person instead, rather than guess.
- I choose the provider and settings with you, and document what is stored, where and for how long.
Draft copy, to confirm
What you get out of it
Visitors get accurate answers to common questions at any hour, the assistant admits what it does not know, and anything important reaches a person.
Who it’s for
- You answer the same customer questions many times a week.
- You have a good library of service pages, FAQs or documents that an assistant could draw on.
- You want visitors helped outside office hours, with a clear route to a person when it matters.
If visitors mostly struggle to find information, better navigation and site search may help more than a chatbot; website redesign and UX work cover that. If the work happens behind the scenes rather than in conversation with visitors, see AI automation and integration.
What’s included
Question review
The questions people actually ask, collected from your inbox, forms and calls, so the assistant is built for them.
Approved knowledge
The pages and documents the assistant may answer from, checked for accuracy and kept up to date.
Boundaries
Topics it must not answer, wording it must not use, and the point at which it offers a person instead.
Human handover
A clear route from the chat to your team, by form, email or a booking, with the conversation passed on where you agree it.
Privacy and consent
What the chat collects, where it is stored, how long it is kept and how that is explained to visitors.
Evaluation and monitoring
A test set of real questions run before launch, and a regular review of conversations to catch wrong or missing answers.
How it works
Collect
We gather the questions your customers ask and the content that already answers them.
Configure
I set up the assistant on its approved content, with its boundaries, handover and privacy settings.
Test
We run the agreed questions, including awkward and off-topic ones, and fix what it gets wrong.
Launch and review
It goes live with monitoring, and we review real conversations after the first weeks to adjust it.
Questions people ask
Draft answers, to confirm
Will the chatbot make things up?
Any AI model can produce a wrong answer. I reduce the risk by limiting it to approved content, testing it on real questions and setting it to hand over when unsure. Monitoring after launch catches what testing missed.
Where are the conversations stored?
That depends on the provider we choose. I document it before launch, so your privacy notice can say what happens to visitors' messages.
Are there running costs?
Usually, yes. Chatbot and AI providers charge you directly, often by usage or by plan. I explain the likely costs before we build.
Can it book appointments or take payments?
It can pass visitors to your existing booking or payment pages. I do not let an assistant make bookings or payments on its own without a person or a tested system in between.
Tell me what you need to achieve.
We can work out the right next step before committing to a solution. You can book this on its own. An ongoing partnership is optional, never a requirement.
Need continuing responsibility rather than a defined project? Ongoing partnerships